pcp
[Top] [All Lists]

Re: [pcp] sheet2pcp: better error handling

To: myllynen@xxxxxxxxxx
Subject: Re: [pcp] sheet2pcp: better error handling
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Thu, 17 Dec 2015 18:29:26 -0500 (EST)
Cc: pcp developers <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <56727776.9080701@xxxxxxxxxx>
References: <56727776.9080701@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: wV2DGBGCPFNzUjnHBxRMKWaN1BHvaw==
Thread-topic: sheet2pcp: better error handling
Hi Marko,

----- Original Message -----
> Hi,
> 
> while testing sheet2pcp for the first time I noticed the error messages
> and error handling could be improved a bit. The below patch gives clear
> indication of a) wrong file name / type used and b) invalid datetime
> used (otherwise will see those "Use of uninitialized value" messages
> and timestamp for all records in the archives will be the Epoch).

Merged, with minor tweak to show a bit more detailed diagnostics.  I also
added to qa/372 - could you double-check that (commit 14d8dabd7) and make
sure it covers this fix sufficiently for the problems you saw?

> There's still one case which I saw but I'm not sure how to best address
> it, namely if you mistakenly have something like this in the map file:
> 
> <metric pmid="60.1.1" indom="60.0" units="1,0,0,PM_SPACE_KBYTE,0,0"
> type="PM_TYPE_U64" sem="PM_SEM_INSTANT">mem.util.used</metric>
> 
> That is, indom defined for a singular metric, then you'll get:
> 
> pmiGetHandle: failed to create handle for metricspec mem.util.used: No error
> 
> Perhaps the earlier pmiAddMetric() error handling should be revisited
> to have more descriptive error messages.

Mmm, yeah, sounds a good idea - the "No error" message might even be an
outright bug in itself?

cheers.

--
Nathan

<Prev in Thread] Current Thread [Next in Thread>